Agent memory design question: what deserves a place in permanent memory?
Miserable-Shock3552 · reddit · 2026-09-22
A developer building a kids' bedtime app with continuing stories — same world and characters, ten episodes per season — asks how to keep episode 8 consistent with episode 1.
His current setup uses two memory layers: a small set of fixed world facts (names, traits, relationships, places) and short per-episode summaries; each new episode receives the facts plus recent summaries, not full history.
The recurring failure: a minor detail from episode one sits in the facts layer and resurfaces in every later episode, since the model treats everything in that layer as important. He's weighing weighting, decay, or a separate "minor details" tier, and asks how others decide what earns permanent memory versus what should fade.
More from coding & agent
- Debate: will all mainstream software be formally verified by 2030? Legacy debt says no — luislamb · 2026-09-22
- OpenMuse: open-source AI coworker that browses the web in a disposable Linux desktop — aniketmaurya · 2026-09-22
- Turn detection experiment: Jev hits 16/16 while tiny-BERT Laya flunks on own data — tobowers · 2026-09-22
- What engineers check before adding a new LLM provider — Rama_Surasani_ · 2026-09-22
- Builder: AI productivity gains in software are 'truly unbelievable' — omnivaughn · 2026-09-22
- OpenAI case study: V7 Go gives enterprise agents long-term memory via Context Graph — xiaohu · 2026-09-22